Chapter 5 Flashcards
What is one problem with facial detection on a computer?
Can recognize a mismatch between faces if the faces are straight on… much harder if they are angled. They are also fooled by makeup.
What was one attempt at having better computer vision?
Spatial layout systems- get okay results, ofte they aren’t so good. Can find borders well but can’t find objects.
Why can’t spatial layout systems find objects?
1) Image clutter-many objects are partially occluded by others. 2) Object variety- there are tons of different versions of the same thing (trees!)
3) variable views- different retinal images depending on our viewpoint, object orientation, lighting conditions etc
4) retinal stimulus is ambiguous (inverse projection)
What is the inverse projection problem?
The task of determining the object that caused that particular image on the retina-impossible to solve using only bottom-up perception!
What was Stephen Palmer’s experiment?
Asked participants to name line-drawn objects, but first they were shown a scene (told to ignore it). Scene was appropriate or inappropriate for the object, and there were also 2 types of objects (ones that look similar and ones that don’t)
What were the results of Stephen Palmer’s study?
Performance (object identification) was faster and more accurate after seeing appropriate context (80% versus 40%) . Also, if the object was presented briefly, people mistake the incorrect, but similar object for the appropriate one.
What happened in the follow-up study to Palmer’s experiment that asked peole to draw the object they saw?
It ends up looking more like the object appropriate to the scene.
What are top-down models of perception?
Perception is influenced by existing knowledge, goals, expectations.
What are the steps of top-down perception?
Expectation—> formulate perceptual hypothesis—->Examine features to check hypothesis—-> recognize stimulus. (ex: seal trainer versus ballroom dancer demo)
How does context work in top-down perception?
The context in which something appears guides how we will perceive it (ex: ink blots over certain letters of a word, examine other letters to decide what is MOST likely.
What are the 3 parts of top-down perception?
1) context
2) attention
3) organization and structure.
What is the jumbled word effect? (context)
Ability to read words in sentences despite having mixd up letters in the middle of words
What is a perceptual set? (context)
Readiness to perceive things in a certain way (ex: Harold Kelly’s guest lecturer, blind date).
What are perceptual constancies?
Tendency to experience the same, stable perception despite changes in sensory input (ex: size constancy, colour constancy, brightness constancy)
What is colour constancy?
Colour is not determined solely by wavelengths of light entering the eye-colour is perceived as constant across different lighting
What is brightness constancy?
When we subtract out a shadow and see something as being the same brightness across different lighting.
What is inattentional blindness? (attention)
Failure to register unattended stimuli in consciousness (ex: Gorilla video)
What kinds of things grab our attention best?
Things of evolutionary importance (threatening faces, sexual things)